The long-term objective of the proposed work is to understand variability in speech intelligibility, with the ultimate goal of developing speech intelligibility enhancement techniques for a wide range of talkers, listeners and communicative situations. The current objective is to explore variability in speech intelligibility through both listener adaptation to the talker and talker adaptation to the listener. In particular, we focus on the special case of speech communication between all possible combinations of native and non-native English talkers. This focus is in-line with current national and global trends towards increasing contact between native and non- native English speakers, and therefore stands to make both theoretical and practical contributions. Our central claim is that variability in overall speech intelligibility is a function of talker-listener sound structure alignment, which can be adjusted in a bi-directional, dynamic manner according to the current communicative conditions. We hypothesize that a mismatch of language background between interlocutors is a bi-directional source of speech intelligibility variability, as well as of cognitive-linguistic innovation. Two predictions of this hypothesis are: (1) variability in intelligibility is related to talker-listener alignment rather than a simple function of talker and/or listener target language proficiency (bi-directional intelligibility variability), and (2) both native and non-native speakers exhibit speech perception and production changes in response to exposure to native and non-native speech (bi-directional innovation). The specific aims are: Aim 1: To develop a large corpus of speech produced by native and non-native speakers of English that includes both scripted and spontaneous, dialogue-based speech samples. The speakers will be carefully selected to cover a range of native language backgrounds and levels of English proficiency. Moreover, in the dialogue portion of the corpus, talkers will be paired in a principled manner, covering 4 different conversation pair types: 2 native talkers, 1 native and 1 non-native talker, 2 non-native talkers from the same language background, and 2 non-native talkers from different language backgrounds. The corpus will be fully transcribed and partially phonetically aligned, creating a valuable resource for the speech research community. Aim 2: To use this corpus to examine bi-directional talker-listener alignment at the level of speech production and perception. In a series of 3 planned experiments we will (i) examine foreign-accented speech intelligibility in relation to inter- and intra-talker acoustic-phonetic variability, (ii) compare overall communicative efficiency across all 4 types of conversation pairs, and (iii) compare the direction and extent of phonetic convergence in dialogues between conversation partners that vary with respect to their levels of proficiency in the target language (English) and in terms of their matched or mismatched native language backgrounds.